It wasn’t something I said that set that off was it?

“Dicoogle is a Open Source project that aims (for now) to index DICOM repositories. Thus, you can search in every field that the image contains. Moreover, Dicoogle is a distributed repository. You can have several repositories in your intranet and they communicate with each other, creating a federate view of the repository.”

This appears to be a sensible approach to the problem asserted to exist and to be solved by sucking the whole country’s medical images into a single separate and of course new and additional storage system.

It is an easy and well-enough understaood problem that I expect the implementation to be adequate, and the design means it fails gracefully, of course.

And is philosophically satisfactory.

Its fundamental principle[2] may[3] be worth adopting. NPfIT started from 0/2 on that basis, and if anything got worse.

So the internet works by providing wires on which sit a load of computers (referred to as “hosts” or “peers”[4] ). Each can talk to each using a simple set of rules. Some vaguely central directories of machines are kept, and anyone may make a list of machines they trust and identifiers for them. This is a task which governments could do well, if they could do it well.

If the computer in the casualty department wants to know something about a patient, then it can send a question to the computer in the patient’s general practice. It could ask around to the Darzi Centre, the Nuffield[5], and the Orthopaedic clinic in Val Thorens[5] as well.

And then the patient should get an account showing what access has been made to their medical record[6], which tends to keep everyone honest, or at least visible.

(Adding a registry and security to this is not as hard as it may seem. Not trivial, and not something to do with secret programs, but the problems are known[8])

[0] After Al Gore steered funding for it through Congress the nerds just got on with making it work.

[1] Correlation does not prove causation

[2] Rough consensus and _running code_

[3] I mean is, but this is supposed to look more tactful

[4] An indication of the attitude which works, and prevails, and is opposite to that favoured in NPfIT thinking, where _our_ machines are regarded as clients of _their_ servers

[5] examples of places excluded by the currentfailed and abandoned approach to NHS sharing of records[6].

[6] whatever a medical record is, something which Accenture, Fujitsu, and CSC turned out to be no more sure about than the DoH, I think. I’m not sure about it either, but I’m quite keen on saying so because I think it is interesting to consider.

[7] refer to note 7 if you want to demonstrate having paid attention

[8] Some people may say understood and solved, but at least the latter group tend to be selling secret source programs that they claim solve them. The former group may be optimistic, but are likely to say that closed source solutions are unreliable, and are correct in that.

It can be patched to make it usable, probably, but it may actually be better to separate its various functions.

Computers permit this without a perceptible increase in errors or effort or cost.

A handwritten addendum greatly increases cost, greatly increases the potential for error, is rarely designed well, accentuating those last faults, scales extremely poorly, and is very expensive.

Handwritten addenda are more common and likely to become more so.

Some of them could be substituted for by carrying a printer and producing sticky labels, as done in 1984 for such tasks as intravenous feeding regimes in some places with an Epson carry-able computer.

The problem, and the solution are likely to be unappreciated by those focused upon hospitals, which regrettably includes those in charge of district nursing and nursing home regulation. They lack the background. They may be educable.

Event-based financial system used, holding state needed.

The problem is that the FP10 form is overloaded[1]. It is an accounting token, a workflow token for pharmacy, a set of instructions to a patient, and it is not, and never has been, a complete list of what the patient should currently be taking, and how.
It is also used as an indication of what GPs prescribe, a task for which it is of course the least good record.[2]

A solution is to automagically produce a state document – IE each time a change of medicines occurs,

This is something which should come via doctor-driven development of IT systems, and introduction of it without IT support – automation – should be opposed vicigorously

[1] it has more than one function.

[2]If you substitute for “prescribe” something about what pharmacists admit to having dispensed for patients then it is marginally more accurate, but it does reflect as well as could be done the prescribing intentions of GPs which oddly is what the people looking at these analyses claim to be or should be trying to understand and sometimes influence.)

Nearly 800 GP practices in England were unable to work after losing access to their patient care records systems, following a series of server failures.”
Health Service Journal

(See earlier entry here: http://defoam.net/wordpress/2011/08/18/general-practice-central-computer-sytems-broken-today-emis-down-across-england/

That is SNOMED plus the Clinical Terms Thesaurus (CTT)

CTT was the later name for the Read Code version 3, or actually Read 3.1, since we[1] regarded 3.0 as no good, largely as it didn’t contain the codes and rubrics from the earlier sets – Read 4 byte and Read 5 byte (most of you are using the latter, EMIS users will remember the former, and the pain of change).

The Mayo Clinic was contracted to handle the combination of the multiaxial SNOMED set with the Read 3.1/CTT. The NHS and Treasury[4] convened a working group to advise them whether a way of determining if this work was done well could be produced. We[2] gave a clear answer. The answer was “no”, but I expect it was done adequately well.

Read/CTT is good at general practice stuff, less good at hospital stuff. SNOMED is good at pathology, and I assume much of the hospital stuff apart from that. Read/CTT already embodied – actually enveloped – the ICD and the UK extensions to ICD 9 and I presume 10[3]

The two sets have an area of overlap, where codes should be mapped onto each other, and areas that don’t overlap where the result will be that you can code a wider range of concepts with the single system.

I think it is more complex, centralised, prescriptive and unevolutionary[5] than is ideal, but that the persistent efforts to hold copyright on these collections of terms and the manner of their handling compels that. And therefore coding systems, as with natural language, should not be restrictively licenced and should be presented as Open Source or Commons.

So it has been planned all this century, or at least intended.

[1] the specialty working group for quality assurance in Clinical Terms/Read/Thesaurus. We didn’t write them, we did criticise and accept.

[2] Just after the Hammersmith train crash, and just about under the line it happened on. I was asked to attend. One member got a curious smile when I remarked the whole lot should be Open Source, because he was unable to announce that his unit at the University of Manchester were about to do just that, until the following week.

[3] it was a while ago, and I’ve not been involved (I’m not sure if any doctors have been) for quite some time.

[4] It was an effort to explain to the Treasury chap why this mattered. I think it was accomplished – two cultures and all that.

[5] as in the evolving nature of language, with loan words and so on

Wildly wrong in the first sentence.

I’m interested in who briefed them that “currently GPs use paper records”. This GP doesn’t, and hasn’t this century. Other bits of the NHS have carefully avoided learning from us, and are setting up systems similar to those we had in 1980, and I suspect the vendors and champions of those wish to compel us to abandon ours and move to theirs.

The record of central NHS IT projects is appallingly bad, but the failed model of development and control has been applied over and over again, and currently the centralisers are trying to take over the only success stories.

Bad.